Update.

diff --git a/main.py b/main.py
index 784474f..0c2ff24 100755 (executable)
--- a/main.py
+++ b/main.py
@@ -1,4 +1,4 @@
-!/usr/bin/env python
+#!/usr/bin/env python
 
 # Any copyright is dedicated to the Public Domain.
 # https://creativecommons.org/publicdomain/zero/1.0/
@@ -511,7 +511,7 @@ class TaskPicoCLVR(Task):
 
         image_name = os.path.join(args.result_dir, f"picoclvr_result_{n_epoch:04d}.png")
         torchvision.utils.save_image(
-            img / 255.0, image_name, nrow=nb_per_primer, padding=1, pad_value=1.0
+            img / 255.0, image_name, nrow=nb_per_primer, padding=1, pad_value=0.0
         )
         log_string(f"wrote {image_name}")
 
@@ -622,15 +622,27 @@ class TaskMaze(Task):
     def compute_error(self, model, split="train", nb_to_use=-1):
         nb_total, nb_correct = 0, 0
         count = torch.zeros(
-            self.width * self.height, self.width * self.height, device=self.device, dtype=torch.int64
+            self.width * self.height,
+            self.width * self.height,
+            device=self.device,
+            dtype=torch.int64,
         )
-        for input in task.batches(split, nb_to_use):
+        for input in tqdm.tqdm(
+            task.batches(split, nb_to_use),
+            dynamic_ncols=True,
+            desc=f"test-mazes",
+        ):
             result = input.clone()
             ar_mask = result.new_zeros(result.size())
             ar_mask[:, self.height * self.width :] = 1
             result *= 1 - ar_mask
             masked_inplace_autoregression(
-                model, self.batch_size, result, ar_mask, device=self.device
+                model,
+                self.batch_size,
+                result,
+                ar_mask,
+                progress_bar_desc=None,
+                device=self.device,
             )
             mazes, paths = self.seq2map(result)
             path_correctness = maze.path_correctness(mazes, paths)
@@ -705,6 +717,7 @@ class TaskMaze(Task):
                 target_paths=paths,
                 predicted_paths=predicted_paths,
                 path_correct=maze.path_correctness(mazes, predicted_paths),
+                path_optimal=maze.path_optimality(paths, predicted_paths),
             )
             log_string(f"wrote {filename}")
 

diff --git a/maze.py b/maze.py
index 81afcd9..fd0a1d2 100755 (executable)
--- a/maze.py
+++ b/maze.py
@@ -146,8 +146,16 @@ def mark_path(walls, i, j, goal_i, goal_j, policy):
         assert n < nmax
 
 
+def path_optimality(ref_paths, paths):
+    return (ref_paths == v_path).long().flatten(1).sum(1) == (
+        paths == v_path
+    ).long().flatten(1).sum(1)
+
+
 def path_correctness(mazes, paths):
-    still_ok = (mazes - (paths * (paths < 4))).view(mazes.size(0), -1).abs().sum(1) == 0
+    still_ok = (mazes - (paths * (paths != v_path))).view(mazes.size(0), -1).abs().sum(
+        1
+    ) == 0
     reached = still_ok.new_zeros(still_ok.size())
     current, pred_current = paths.clone(), paths.new_zeros(paths.size())
     goal = (mazes == v_goal).long()
@@ -214,6 +222,7 @@ def save_image(
     score_paths=None,
     score_truth=None,
     path_correct=None,
+    path_optimal=None,
 ):
     colors = torch.tensor(
         [
@@ -276,16 +285,26 @@ def save_image(
         )
         imgs = torch.cat((imgs, c_score_paths.unsqueeze(1)), 1)
 
+    img = torch.tensor([224, 224, 224]).view(1, -1, 1, 1)
+
     # NxKxCxHxW
-    if path_correct is None:
-        path_correct = torch.zeros(imgs.size(0)) <= 1
-    path_correct = path_correct.cpu().long().view(-1, 1, 1, 1)
-    img = torch.tensor([224, 224, 224]).view(1, -1, 1, 1) * path_correct + torch.tensor(
-        [255, 0, 0]
-    ).view(1, -1, 1, 1) * (1 - path_correct)
+    if path_optimal is not None:
+        path_optimal = path_optimal.cpu().long().view(-1, 1, 1, 1)
+        img = (
+            img * (1 - path_optimal)
+            + torch.tensor([0, 255, 0]).view(1, -1, 1, 1) * path_optimal
+        )
+
+    if path_correct is not None:
+        path_correct = path_correct.cpu().long().view(-1, 1, 1, 1)
+        img = img * path_correct + torch.tensor([255, 0, 0]).view(1, -1, 1, 1) * (
+            1 - path_correct
+        )
+
     img = img.expand(
         -1, -1, imgs.size(3) + 2, 1 + imgs.size(1) * (1 + imgs.size(4))
     ).clone()
+
     for k in range(imgs.size(1)):
         img[
             :,